梨和苹果糖度在线检测通用数学模型研究

发布时间：2018-06-19 11:49

本文选题：在线检测 + 可溶性固形物　；参考：《光谱学与光谱分析》2017年07期

【摘要】：采用可见/近红外光谱技术在线检测水果糖度,每个水果品种要单独建模,模型升级维护耗时费力。探讨建立苹果、梨等薄皮水果可溶性固形物(SSC)在线检测通用数学模型的可行性。利用自行设计的可见/近红外漫透射光谱在线检测系统,在积分时间80ms、单线速度5个/s的条件下,采集新梨7号、砀山酥梨、玉露香梨和富士苹果四种水果的可见/近红外漫透射光谱。分析了四种水果的可见/近红外漫透射光谱响应特性,采用变异系数法和连续投影算法,筛选通用数学模型建模用光谱变量,并建立了偏最小二乘和最小二乘支持向量机梨与苹果梨通用数学模型。采用新样品评价模型的预测能力,变异系数法筛选光谱波段建立的偏最小二乘通用数学模型预测精度最高,通用模型预测梨和苹果梨模型预测均方根误差分别为0.49%和0.55%,通用模型预测相关系数分别为0.88和0.93;独立模型预测新梨7号、玉露香梨、砀山酥梨和富士苹果的预测相关系数分别为0.93,0.91,0.88和0.95,预测均方根误差分别为0.40%,0.42%,0.41%和0.46%。通用数学模型的预测精度略低于每个品种的独立数学模型,但是通用模型的通用性高于单一模型。实验结果说明采用变异系数法结合偏最小二乘法建立薄皮水果在线检测通用数学模型,实现四种水果糖度在线检测是可行的。
[Abstract]:Using visible / near infrared spectroscopy (NIR) technology to measure the degree of fructose on line, each fruit variety needs to be modeled separately, and the model upgrading and maintenance takes time and effort. To explore the feasibility of establishing a general mathematical model for on-line detection of soluble solids in apple, pear and other thin-skinned fruits. Using the self-designed on-line detection system of visible / near infrared diffuse transmission spectrum, under the condition of integrating time 80 Ms and single line velocity 5 / s, Xinli 7 and Dangshan pear were collected, Dangshan pear, Dangshan pear, The visible / near infrared diffuse transmission spectra of four kinds of fruits, Julu pear and Fuji apple. The response characteristics of visible / near infrared diffuse transmission spectra of four kinds of fruits are analyzed. The spectral variables used in the modeling of general mathematical models are selected by means of variation coefficient method and continuous projection algorithm. The general mathematical models of pear and apple pear based on partial least squares and least squares support vector machine are established. Based on the prediction ability of the new sample evaluation model, the partial least squares general mathematical model established by the coefficient of variation method for spectral band selection has the highest prediction accuracy. The root-mean-square error (RMS) of the two models was 0.49% and 0.55%, respectively, and the correlation coefficient was 0.88 and 0.93.The independent model was used to predict Xinli 7, Yulu fragrant pear. The predictive correlation coefficients of Dangshan pear and Fuji apple were 0.93 ~ 0.91 ~ 0.88 and 0.95, respectively. The root mean square error was 0.400.42% and 0.46%, respectively. The prediction accuracy of the general mathematical model is slightly lower than that of the independent mathematical model of each variety, but the general-purpose model is more general than the single model. The experimental results show that it is feasible to establish a general mathematical model for on-line detection of thin skin fruits by means of coefficient of variation method combined with partial least square method and to realize the on-line detection of sugar content of four kinds of fruits.
【作者单位】：华东交通大学机电与车辆工程学院;
【基金】：国家自然科学基金项目(61640417) 南方山地果园智能化管理技术与装备协同创新中心(赣教高字[2014]60号) 江西省优势科技创新团队(20153BCB24002)资助
【分类号】：O657.3;TS255.7

【相似文献】